Home // DATA ANALYTICS 2021, The Tenth International Conference on Data Analytics // View article

A Comparison of Machine-Learned Survival Models for Predicting Tenure from Unstructured Résumés

Authors:
Corné de Ruijt
Vladimer Kobayashi
Sandjai Bhulai

Keywords: Human resource management; turnover prediction; résumé mining; machine-learned survival models; job churn

Abstract:
This paper explores to what extent job seekers' future job tenures can be predicted using only the information contained in their own résumés. Here, job tenure is interpreted as the time spent in a single job occupation. To do so, we compare the performance of several machine-learned survival models in terms of multiple error measures, including the Brier score and the C-index. The results suggest that ensemble methods, such as random survival forest and Cox boosting, work well for this purpose. We further find that in particular time-related features, such as the time a person has already worked in a particular field, are predictive when predicting the person's future tenure. However, the results also show that this prediction task is difficult. There is substantial subjectivity in both how job seekers define their jobs, and at what level of granularity they indicate their job tenures. As a result, the best performing models (survival ensemble methods) only perform marginally better than the used benchmark (a Kaplan-Meier estimate).

Pages: 1 to 6

Copyright: Copyright (c) IARIA, 2021

Publication date: October 3, 2021

Published in: conference

ISSN: 2308-4464

ISBN: 978-1-61208-891-4

Location: Barcelona, Spain

Dates: from October 3, 2021 to October 7, 2021